Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Qualcomm Nuvia vs Apple Mx
- Thread starter crazy dave
- Start date
-
- Tags
- amd apple silicon intel nuvia qualcomm
- Sort by reaction score
Umm…
You do know that is an 8.5 year old computer? My iPhone 12 Pro is faster. I can’t see the relevance.
Still puts into perspective the progress from 130W to 7W that's almost almost twice as fast. Perhaps in the future we'd be comparing 0.7W that's faster than today's 7W.
So Intel or AMD (who uses tsmc) isn’t competing in the (extremely lucrative) mobile space because they don’t feel like it?
Otellini deeply regretting passing on the opportunity when Apple approached Intel be the chip supplier for iPhone.
“Deeply regret“ is an overstatement .
“… At the end of the day, there was a chip that they were interested in that they wanted to pay a certain price for and not a nickel more and that price was below our forecasted cost. I couldn't see it. It wasn't one of these things you can make up on volume. And in hindsight, the forecasted cost was wrong and the volume was 100x what anyone thought." ….”
Paul Otellini's Intel: Can the Company That Built the Future Survive It?
As the CEO steps down, he leaves the Intel machine poised to take on the swarming ecosystem of competitors who make smartphone chips.
this is more “ if a , would a , should a “. Mixed in with trying to justify buying a discrete cellular modem company ( Infineon) just as Apple dumped them as a supplier .
Three major things .
1. the forecast was wrong . There is a difference between good decisions and good outcomes . if Apple was flexible about price they may have done it. If the cost projection were lower may have done it. The volume that Apple later did was immaterial if Intel could deliver on costs . It would be high volume stuff they made no money on . With information he had , there should not be regrets in that decision. Apple was not trying to dump a giant pile of money into intel's pockets .
2. Intel was making SoCs that Palm and others were using . What Apple was asking for probably would need Imagination Tech GPU included. Also don’t think StrongArm on the leading edge node.
Intel didn’t have a good foundry business set up and this would need significant IP working inside and fast. Was the internal politics of the x86 group going to let Apple cut in-line before them for leanest node and discount prices ?
[ the forecast suppposely being wrong could have several factors contributing . ]
3. When Apple started off with Samsung , they immediately started building an in house design . Would have been any different for Intel ? By time Otellini did this exit interview Apple had dumped Samsung as designer and were working on dumping Imagination Tech . If Intel wasn’t a full service fab for independent IP then would not have the leverage that Samsung had to hang onto at least the fab part of Apple’s orders for a couple more years .
Samsung being Apples SoC didn’t lead long term to Samsung being biggest player in Smartphine SoC market . Would Intel be in better shape given there inter bias for x86 solutions? . Intel didn’t just drop ball on Apple . They chucked the whole StrongArm operation .
Apple has the margins to get away with modem less SoC . Even if Intel has a better than average SoC by time Apple dropped them if didn’t have world class modem Qualcomm still would hav3crun then over.
Otellini’s biggest regret should have been letting Krzanich anywhere near the COO or CEO chair. A initial glitch 14nm and outright stumble 10nm was on Krzanich watch ( was leader of fab before COO. So problem shouldn’t have snuck up on him )
Next up would have been buying discrete modem just as discrete modem died in overal smartphone market. At that point chasing “Wall Street” hype about how to recover from missing out on iPhone so tried to grab what thought would be other high margin component . [ Infineon slow dragging on 4G and not having a “one world” modem meant Intel paid hefty premium for not much. ].
I think he’s had plenty of time to think about how to answer the question “properly”. He has to make it sound like it would have been a money losing proposition and ‘no one would have taken the deal’. Truth is, we have no idea how much was offered and I doubt anyone at Apple’s even interested in refuting his claim. We also know that contracts can be written such that, in the out years, they can renegotiate for better terms. He wants everyone to believe that Apple wanted an “in perpetuity” contract for a financially ruinous amount.
If, in the intervening years, Intel had NOT tried, over and over again, to enter that market, I’d say there’d be little to no regret. But, looking at Apple’s volumes in a market they tried to enter and failed AND looking at the public migration FROM Intel, if he says anything like, “I wouldn’t have done anything different,” he’s not telling the truth.
And, in the very next passage, just after your quote.
That sounds like prettttty deep regret, there.
New Qualcomm interview with the Senior Director for PCs and Ian @ Anandtech:
His answer was much more in line with my thinking wrt Apple, the M1, Intel, Qualcomm, and Windows:
Another interesting part:
So they were just a performance core, maybe a touch of fabric (bit vague there but it sounds like the fabric is mostly Qualcomm or newly developed with Qualcomm), are probably having to develop new efficiency cores, and the GPU will be Adreno-based (not surprising).
His answer was much more in line with my thinking wrt Apple, the M1, Intel, Qualcomm, and Windows:
Another interesting part:
So they were just a performance core, maybe a touch of fabric (bit vague there but it sounds like the fabric is mostly Qualcomm or newly developed with Qualcomm), are probably having to develop new efficiency cores, and the GPU will be Adreno-based (not surprising).
Qualcomm are vastly superior in implementing their SOC's with ARM reference CPU core designs than MediaTek and Samsung both, we know they've acumen in that regard. And hell, Qualcomm's A77's in the 865 on TSMC N7, @ 2.65-2.8GHz are notably ahead of the the Dimensity 1200's A78 on N6 @ 3GHz in energy efficiency from SpecInt2017.
Even the 8 Gen 1 X2 core on garbage Samsung 4NM LPE tops the chart for non-Apple core energy efficiency, and while the Dimensity 9000's TSMC N4 X2 is 15-20% more efficient, I suspect that margin of advantage will turn toward Qualcomm's favor with the Gen 8 + on N4, given the reasonable expectation from almost evey piece of data we have that QC has IP and talent that renders their implementations of these cores more effective in some way or another.
More importantly QC were fairly direct about about Adreno 730's performance and efficiency upgrades gains they claimed back in November of this past year (2021), and while the current 8 Gen 1 sporting this GPU is on Samsung 4NM LPE, which as noted is pathetic, they've essentially got themselves back in the game with it. Both S22 Plus and S22 Ultra are neck-and-neck with the iPhone 13 Pro.
They will scale Adreno up undoubtedly as Apple did with their graphics, and with a better process that is even vaguely comparable to Apple's TSMC N5-derivatives today or in the near future, I find it plausible Qualcomm will come out ahead by a decent margin on power efficiency and performance in GPU's. Remember, Qualcomm took Apple to the woodshed on GPU's until the A12, and while they deserve credit for that coup (Apple), I think we are going to see a return to form for The Modem Company here soonz
As for the wishcasting about Nuvia from various Apple Disciples here: Qualcomm wouldn't buy Nuvia for the cores to come out worse than the reference cores, and even if they took the reference cores and clocked them higher replete with ample cache on a decent node, they'd have a laptop SOC capable of competing with AMD and Intel by way of high-density library choices and IPC (Zen 3 is not on par with the X2 lol) but they're going further than that, so...
The only reason the 8CX Gen 3 has the X1's and A78's instead of tbe X1's and A710's is that Microsoft did not yet support Arm V9. If they did, then even on Samsung 4NM LPE with those X2's, they'd have a machine capable of much more.
Between Nuvia, Adreno gains, a TSMC return and their modem/AI prowess + Microsoft's recent uptick in WoA interest, I think there is a deluge of undeserved doom about this. I'll be happy to admit I'm wrong, but I bet the Nuvia cores will be more energy efficient than the reference X3 and A720, and have higher peak throughput than the former for the (if they go heterogenous) big core. I'd wager the performant core will be within +-10% of the current Apple Firestorm core in performance and plausibly more energy efficient. I'd also wager Adreno is going to increasingly make some heads turn on a half-decent node. Betting on +-25% of M2 GPU efficiency.
Last image, the Spec graph, shows Qualcomm beating the Dimensity 1200's 3GHz A78's on N6 with... their 865's Prime A77 core sporting less frequency and *technically* inferior process node (N7) afaict.
Said Spec2017 graph also demonstrates even with Samsung and the X2, their Snapdragon 8 Gen 1 isn't that bad on energy efficiency. Assuredly better than most any Ryzen core or Intel core lol.
Lastly, Adreno 730 is the real deal. They didn't lie at the investor conference.
Even the 8 Gen 1 X2 core on garbage Samsung 4NM LPE tops the chart for non-Apple core energy efficiency, and while the Dimensity 9000's TSMC N4 X2 is 15-20% more efficient, I suspect that margin of advantage will turn toward Qualcomm's favor with the Gen 8 + on N4, given the reasonable expectation from almost evey piece of data we have that QC has IP and talent that renders their implementations of these cores more effective in some way or another.
More importantly QC were fairly direct about about Adreno 730's performance and efficiency upgrades gains they claimed back in November of this past year (2021), and while the current 8 Gen 1 sporting this GPU is on Samsung 4NM LPE, which as noted is pathetic, they've essentially got themselves back in the game with it. Both S22 Plus and S22 Ultra are neck-and-neck with the iPhone 13 Pro.
They will scale Adreno up undoubtedly as Apple did with their graphics, and with a better process that is even vaguely comparable to Apple's TSMC N5-derivatives today or in the near future, I find it plausible Qualcomm will come out ahead by a decent margin on power efficiency and performance in GPU's. Remember, Qualcomm took Apple to the woodshed on GPU's until the A12, and while they deserve credit for that coup (Apple), I think we are going to see a return to form for The Modem Company here soonz
As for the wishcasting about Nuvia from various Apple Disciples here: Qualcomm wouldn't buy Nuvia for the cores to come out worse than the reference cores, and even if they took the reference cores and clocked them higher replete with ample cache on a decent node, they'd have a laptop SOC capable of competing with AMD and Intel by way of high-density library choices and IPC (Zen 3 is not on par with the X2 lol) but they're going further than that, so...
The only reason the 8CX Gen 3 has the X1's and A78's instead of tbe X1's and A710's is that Microsoft did not yet support Arm V9. If they did, then even on Samsung 4NM LPE with those X2's, they'd have a machine capable of much more.
Between Nuvia, Adreno gains, a TSMC return and their modem/AI prowess + Microsoft's recent uptick in WoA interest, I think there is a deluge of undeserved doom about this. I'll be happy to admit I'm wrong, but I bet the Nuvia cores will be more energy efficient than the reference X3 and A720, and have higher peak throughput than the former for the (if they go heterogenous) big core. I'd wager the performant core will be within +-10% of the current Apple Firestorm core in performance and plausibly more energy efficient. I'd also wager Adreno is going to increasingly make some heads turn on a half-decent node. Betting on +-25% of M2 GPU efficiency.
Last image, the Spec graph, shows Qualcomm beating the Dimensity 1200's 3GHz A78's on N6 with... their 865's Prime A77 core sporting less frequency and *technically* inferior process node (N7) afaict.
Said Spec2017 graph also demonstrates even with Samsung and the X2, their Snapdragon 8 Gen 1 isn't that bad on energy efficiency. Assuredly better than most any Ryzen core or Intel core lol.
Lastly, Adreno 730 is the real deal. They didn't lie at the investor conference.
Attachments
Last edited:
All this conversation with Qualcommm and they keep mentioning ARM and Windows... Worries me about the future of Linux on ARM.New Qualcomm interview with the Senior Director for PCs and Ian @ Anandtech:
His answer was much more in line with my thinking wrt Apple, the M1, Intel, Qualcomm, and Windows:
Another interesting part:
So they were just a performance core, maybe a touch of fabric (bit vague there but it sounds like the fabric is mostly Qualcomm or newly developed with Qualcomm), are probably having to develop new efficiency cores, and the GPU will be Adreno-based (not surprising).
Especially since both Intel and AMD have been contributing a lot to the public Linux kernels and dropping some good driver support within the first 3 months of a product release..
There are already a gazillion affordable SBCs running Linux on ARM plus Qualcomm is after the corporate money hence the focus on Windows on ARM.
I'm sure QC will have some support for Linux, mostly I'm sure they would love to have their chips inside Chromebook. The bigger problem is their driver support. They're going to need to support PCs a lot longer than they currently support their phone SoCs.
They will undoubtedly have some advantages, but Apple still has the advantage of complete end-to-end integration that nobody else can really match. At least not yet.
They will undoubtedly have some advantages, but Apple still has the advantage of complete end-to-end integration that nobody else can really match. At least not yet.
I would consider myself optimistic on Nuvia delivering a good core, but some caution is *definitely* warranted as there was no tape out just a simulated design when Qualcomm bought them. I think the previous interview or another one said that explicitly.
As others have said already, I wouldn't worry too much. The Qualcomm-Nuvia cores will likely be compatible with basic ARM designs to the point that Linux on ARM will "just work" and the Qualcomm rep here did stress that they consider their platform to be for more than just Windows (mostly talking about Chromebooks). While all OSes require a bit of tweaking to get the most out of a chip, I think a reverse engineering project like Asahi Linux is extremely unlikely to be needed for Linux to work out of the box on these chips next year.
I very much doubt that MS will let them get away with what Google lets them get away with (and a lot of that is on the vendors/OEMs too) - its a different model/ecosystem. I know a lot of people were up in arms about MS' Windows 11 cutting off some older hardware with new requirements, but think of that outcry and what that means about that ecosystem and its expectations and yeah ... I don't foresee the same kinds of issues here.
Last edited:
Yeah the driver support and the GPU support as a whole it's a whole different picture, those are the two major topics for me.
I daily drive my 14" MBP obviously, however I have a Thinkpad and a Dell XPS running Arch and Pop OS respectively. However I'm curious to see how this Qualcomm on PC will behave as a whole especially for distro hoppers like me. (For those curious I daily drive a Mac because for me paired with home-brew and some other apps it's somewhat like Linux but running on top of a high-end hardware using ARM)
But honestly these last 2 years for a Linux lover have been amazing, things as a whole on the x86 land have improved quite dramatically and even more so with Intel/AMD hiring even more Linux engineers and developers to further reinforce that.
We are no longer in the timeline where Thinkpad with 3-4 years being the only ones using Wi-Fi, Bluetooth and etc.. fully working on a Linux Kernel
Sorry for derailing the thread here with the Linux Support...But BACK ON TOPIC.
I'm curious to see Qualcomm offer as a whole especially on the "M1 Pro" playing field essentially because for now I think Apple has this in the bag for the time being.
I don't doubt they can do something between a M1 --> M2, but the way the final product might end up might not be as a great package as a whole as a MacBook Air M1 for example.
Mostly fair RE: Nuvia. I mean they could blow it, their last custom uArch wasn't that great per se. Still, as far as I am concerned, I've heard it's more energy efficient than the X3 and A720, at least, if not truly 2K GB5 @ 3.5 watts or whatever.
Really though, as long as peak throughput is in the GB5 1600-2000 range @ 3-5.5W, it's a win to me. Though I hope that for phones eventually, they eer on the more power efficient side even at detriment to energy efficiency per se, if only because perceptibly hot peak surface heat to the touch sucks.
The Qualcomm manager in the interview mentions Chrome OS . A lot of that is Linux at the foundation level. Ampere and Graviton2 server SoCs are prima all about Linux. Linux isn’t going atrophy and die here. Qualcomm would sell to Linus laptop vendors if they ask For a large order of SoCs.
Technically if the Qualcomm refernce model have a setting on their Pulton boot infrastructure to allow signed boot of another OS then would need much more than that to keep Linux moving forward.
But need to actually sell a substantial number of units to keep Qualcomm interested . Windows sells . ChromeOS sells .
android , chromeOS would keep Linux in contact with Arm even if we’re not direct motivators out there.
all the major cloud services vendors need Linux on Arm . It isn’t going anywhere .
Last edited:
Eh I'm quite confident whatever they ship will beat the original Air in sustained performance (the MBP m1 fans are objectively incredible if behind the former high-end MBP's fans or the novel M1 Pro/Max so that's different)Yeah the driver support and the GPU support as a whole it's a whole different picture, those are the two major topics for me.
I daily drive my 14" MBP obviously, however I have a Thinkpad and a Dell XPS running Arch and Pop OS respectively. However I'm curious to see how this Qualcomm on PC will behave as a whole especially for distro hoppers like me. (For those curious I daily drive a Mac because for me paired with home-brew and some other apps it's somewhat like Linux but running on top of a high-end hardware using ARM)
But honestly these last 2 years for a Linux lover have been amazing, things as a whole on the x86 land have improved quite dramatically and even more so with Intel/AMD hiring even more Linux engineers and developers to further reinforce that.
We are no longer in the timeline where Thinkpad with 3-4 years being the only ones using Wi-Fi, Bluetooth and etc.. fully working on a Linux Kernel
But BACK ON TOPIC.
I'm curious to see Qualcomm offer as a whole especially on the "M1 Pro" playing field essentially because for now I think Apple has this in the bag for the time being.
I don't doubt they can do something between a M1 --> M2, but the way the final product might end up might not be as a great package as a whole as a MacBook Air M1 for example.
Look at it this way. Say Qualcomm ditched their custom cores and let's imagine Nuvia have nothing superior to the reference cores even after Qualcomm's superior implementation of said cores - just write it all off and imagine they have to do things as others do. MediaTek sucks at implementing ARM core IP. Even then, their X2 on TSMC N4 performs at about the same to the Qualcomm X2 on Samsung 4NM LPE in Spec, albeit at 20% less power or so. We're still talking about a core for either of them that's hitting ~ 1250-1275 @ 3GHz or 3.05GHz on Geekbench or about a 4.5-5 on SPECint2017, but at 3.25-3.5W for one and 2.6-3W for the other (MediaTek on N4).
Just get QC to throw it to 3.3GHz (and the X2's cell libraries with TSMC N5 allow for it apparently, up to 3.5 even) and more cache, plus the instant gains in power. Would a core hitting 1500 GB5 ST @ maybe 3.5ish watts be that goddamn bad? Remember the X2's are big cores but not Apple-sized, so that's feasible with a 10% clock boost, shift from Samsung 4NM LPE to N5/N4, and more L3/SLC cache.
And I mean... again.
This is the reference core. I will eat my words should the Nuvia cores, if fabricated on TSMC N5, N5P, N4, N4P, or N3, not realize > 1550 Geekbench or beat the MediaTek Dimensity 9000's X2 (and the Snapdragon 8 Gen 1's X2 IPC for that matter) in IPC and energy efficiency. There's no ****ing way Qualcomm are that goddamn lazy and moronic. In the past, their Windows designs were clearly an afterthought and it's very obvious the acquisition and repeated interviews and updates signal a departure from that seemingly agnostic, aloof stance on their own product. As such I expect the work they will show off to match that signal to a degree, though opinions will vary.
As for Linux: WoA still uses UEFI and ACPI, AFAICT. Linux will eventually be workable or will be an order of magnitude less arduous than it has been for the admittedly impressively dedicated masochists like Hector Martin with Linux on the M1 systems.
And frankly I'm not going to read through every But Linux Post (though I value having UEFI and ACPI and the ability to boot an alternative OS with ease vs the latest Macs) and I'm sure someone else may have mentioned the above.
But really I think the hystericism surrounding this subject is unnecessary, though also entirely unsurprising.
I suspect that would be at least 2-3 years out . They probably don’t want to ignore it forever but more than a good chance they will want to weave these “next get “ cores into their current SoC products first then then move onto new SoC coverage areas . It is way more cost effective to hold onto the customers you have now then to chase new ones.
It is a toss up whether they would be lured by server like options next or by moving up the laptop ladder.
depends upon how want to scale the GPU at least as much as the ’CPU’ application cores. Trading higher average SoC selling price for volume could generate more revenue.
Qualcomm also has a super high hype market to service for VR/AR headsets. if Apple drops a SoC that is serious threat there that would likely prioritize over the bigger laptops and lower end all in ones .
again do not want to loose customers already spent money to get .
Last edited:
No, Arm's objectively balanced and modern v8-v8.4 (and v9 now) ISA is a pretty big contributor to what we've seen in the last few years from a lot of this. Hell, I bet Intel 7's High-Density libraries would prove superior to Samsung 5NM LPE for an Arm core, the only issue is the main core we've seen on Intel 7 is Gracemomt, and Gracemont is.... well the first Atom I don't hate but it's no X2 or even A78/A710.